WO2020164670A1

WO2020164670A1 - Method for trajectory planning of an assistance system

Info

Publication number: WO2020164670A1
Application number: PCT/DE2020/200005
Authority: WO
Inventors: Fabian Becker; Radoy STANCHEV
Original assignee: Conti Temic Microelectronic Gmbh; Continental Automotive Gmbh
Priority date: 2019-02-12
Filing date: 2020-01-16
Publication date: 2020-08-20
Also published as: EP3924797A1; CN113412458A; US20220176958A1; CN113412458B; JP2022515437A; JP7266684B2; DE102019201800A1

Abstract

The invention relates to a method for trajectory planning of a driver assistance system, particularly an assistance system for longitudinal or transverse control, wherein a trajectory (T1, T3-T7) having a total duration (te) which can be set is determined and the trajectory (T3, T5, T7) is divided into segments, wherein each segment has a variable segment duration (Δt1, Δt2, Δt3) and the total of the segment durations (Δt1, Δt2, Δt3) corresponds to the set total duration (te) of the trajectory (T3, T5, T7).

Description

Procedure for planning a trajectory of an assistance system

The present invention relates to a method for trajectory planning of an assistance system for a means of transport or a vehicle, in particular a driver assistance system for longitudinal and / or lateral control, as well as a trajectory planner for carrying out the method according to the invention and an assistance system or driver assistance system in which the trajectory planning takes place by means of the method according to the invention.

Technological background

Modern means of transport, such as B. vehicles, bicycles, motorcycles, planes, Droh NEN, watercraft, boats and the like, are increasingly equipped with assistance systems or driver assistance systems. In particular, in the field of Fahrzeugtech technology, the recognition of road users or other vehicles, pedestrians and the like as well as the detection or assessment of lane markings (z. B. road borders or lane markings) are elementary functions in modern driver assistance systems and are z. B. used in transverse and longitudinally regulating assistance functions (lateral and longitudinal functions), such. B. with an adaptive cruise control (ACC) or automatic distance control (ADR), a lane keeping assistant (LKA, lane keep assist) or an emergency braking assistant (EBA, emergency brake assist). For example, the trajectory to be driven or the movement path of the respective means of locomotion or vehicle can thereby be determined. Using suitable sensors, static targets or objects can be detected, whereby z. B. the distance to a vehicle ahead or the course of the road can be estimated. For object recognition z. B. radar, lidar or camera sensors can be used.

Generic control concepts for assistance systems are based z. B. on trajectory planning by means of optimization. Furthermore, assistance systems of this type often have limited computing resources, e.g. B. on a (radar) control unit, and should therefore be used in conjunction with efficient optimization methods. For example, WERLING (in Werling, Moritz: “A new concept for generating and stabilizing trajectories in time-critical traffic scenarios”. KIT Scientific Publishing, Karlsruhe, 2011) and RATHGEBER (in Rathgeber, Christian: “Trajectory planning and follow-up control for assisted to highly automated driving ”. Technische Universität Berlin, 2016) suggests the analytical solution of a simplified optimal control problem. To do this, in a first step, all restrictions (e.g. jerk and acceleration) are disregarded and the The end time and the final state of the trajectory are assumed to be known. Third to seventh order polynomials result as solutions for the trajectories. Subsequently, the search area spanned by the end time and the final speed or end position is rasterized and a trajectory is calculated for each raster point. A quality measure is calculated for each trajectory. The quality measure is a criterion for evaluating the trajectories, with z. B. the course of the acceleration can be evaluated. In a second step, the trajectories are checked for violations of the restrictions and, if necessary, excluded from the set of valid trajectories. The trajectory with the lowest quality measure from the remaining trajectories is then the result of the optimization. Due to the description of a trajectory by a single polynomial, as with WERLING, the acceleration limits can only be reached selectively. RATHGEBER therefore proposes three-part trajectories: building, holding and reducing the acceleration, each with a polynomial. However, the trajectories for building up and reducing the acceleration have a fixed predetermined duration, which can result in situation-related disadvantages in practice. Furthermore, z. B. for the polynomial-based trajectory planning specific functional extensions of a generic assistance system (z. B. ACC) are not taken into account.

Furthermore, GORJESTANI et al. (in Gorjestani, A .; Shankwitz, C. and Donath, M.: "Impedance Control for Truck Collision Avoidance"; In: Proceedings of the American Control Conference, 2000) a virtual bumper to implement distance control.

State of the art in print

DE 10 2017 200 580 A1 describes a method for optimizing maneuver planning for a vehicle. To carry out the method, the method comprises a planning level which is divided into at least three different abstraction levels for all planning layers of the planning level. A combination of continuous planning and semantic information takes place by grouping several identified maneuver options. In addition, the success of each maneuver option is assessed, taking into account uncertainties in the behavior of other road users, in order to select the best strategy for carrying out the maneuver. task

The present invention is based on the object of providing an improved method for trajek- torien planning of an assistance system and an improved assistance system, in which the disadvantages of the prior art are overcome.

Solution of the task

The above problem is solved by the entire teaching of claim 1 and the claims that are ordered. Expedient embodiments of the invention are claimed in the sub-claims.

According to the invention, in the proposed method for trajectory planning of a driver assistance system, in particular a longitudinal and / or transverse control system (e.g. ACC, ADR, EBA, LKA system or the like), at least one trajectory with a definable total duration is initially determined , which is divided into segments, preferably three segments. Each of the segments has a variable segment duration, the sum of the respective segment durations corresponding to the previously determined total duration of the trajectory, i.e. H. while the individual segment durations are designed to be variable or changeable, the total duration or total length of the trajectory remains unchanged. An expansion of the basic functionality of the respective assistance system can also be made possible in a simple manner, e.g. B. to achieve a specific function extension, such as the predictive speed adjustment when cornering, the predictive Ge speed adjustment for detected traffic signs, the support of the overtaking process through acceleration and / or the prevention of the overtaking process on "slower lane", z. B. right-hand overtaking maneuvers on motorways.

In a simple manner, when the trajectories are subdivided, a first segment can be provided to build up the acceleration, a second segment to maintain the acceleration and a third segment to reduce the acceleration. Furthermore, these segments can also be divided into further sub-segments and / or have segments before, after and / or in between. In particular, the segments for building up and reducing the acceleration should not have a fixed segment duration, so that they can be easily adapted to the respective situation. This improves the flexibility and usability of the entire system to a particular degree. Preferably, the accelerations in the building up and in the declining segment of the trajectory are each described by a third order polynomial for speed control. This has the advantage that such a description or calculation can be implemented particularly easily.

It is particularly useful if the segment duration of the respective segments is determined on the basis of a quality measure. For example, this quality measure can be selected so that it corresponds to the integral component of the quality measure for evaluating a one-part trajectory, which enables, among other things, the direct replacement of a one-part trajectory with a three-part trajectory.

Furthermore, the segment duration of the first segment can be determined as a function of the segment duration of the third segment or vice versa; H. the segment duration of the first segment can e.g. B. be determined via a quadratic equation depending on the segment duration of the third segment.

Appropriately, a three-part trajectory can be calculated for distance control, in which the first and second segments correspond to the segments of the speed control, while the third segment is described by a polynomial of a different order, in particular a fifth order, so that the trajectory transforms the system into the desired Final state (acceleration, speed and position) transferred.

The segment duration of one or more of the segments is preferably selected in such a way that the quality measure of the three-part trajectory is minimal or reduced.

It has proven to be particularly advantageous if a subordinate optimization is provided for selecting the segment duration of one or more of the segments. This further simplifies the selection.

The trajectory can expediently be planned by varying the total duration of the respective trajectory.

Furthermore, an adaptive search space with grid points for determining a trajectory can be provided, the target states of the trajectories in the search space being selected on the basis of a shift of the grid points. In particular because the grid points are shifted towards the optimal trajectory, ie that a concentration of the grid points indicates the optimal trajectory. The adjustment of the grid points is preferably carried out iteratively over several time steps, ie the target points can be varied or adapted using an iterative (over several time steps) procedure.

Appropriately, a spring-damper system or a mass-spring-damper system can be provided, which is used to generate steady and consistent target states (distance, speed and acceleration) of the means of locomotion, which are the target points of the trajectory and / or represent braking motion planning.

According to a special embodiment of the method, a spring-damper system can be arranged as a virtual bumper between the means of locomotion and a means of locomotion moving ahead, in addition to trajectory planning, depending on the situation (e.g. in the area close to a standstill). The dynamics of the virtual bumper can be defined by the definable distance between the means of transport (e.g. host vehicle and vehicle in front), the speed, the acceleration, the mass of the means of transport and / or the (virtual) spring path.

Appropriately, at least one acceleration and / or speed plateau can be provided as a buffer for system deviations that occur. For example, an acceleration plateau can be provided before the means of locomotion comes to a standstill, which specifies the possible target state range for the trajectory planning. Starting from the plateau, the means of locomotion can be defined (or controlled) and brought to a standstill in order to bring about gentle acceleration processes that are familiar to the driver when stopping (e.g. slow braking). This allows z. B. abrupt and unwanted braking maneuvers can be avoided.

A trajectory planner can expediently be provided for determining the trajectory. For example, such a trajectory planner can be designed as a hardware or software module, so that the respective system can easily be pre-assembled at the factory.

In a practical way, the trajectory planner comprises several modules and / or levels. These can e.g. B. permanently configured or modularly interchangeable and / or addable so that the respective functions of the individual modules and levels can be selected user-specific or function-specific. The range of functions and the preconfigurability of the respective trajectory planner are thereby simplified to a particular degree, whereby costs and time can be saved to a particular degree.

Furthermore, the trajectory planner can include a coordination level for the situation and function-specific specification of a target state and a planning level for determining a trajectory based on the target state. Furthermore, the coordination level and / or the planning level can also have a modular structure. For example, the coordination level can include a speed module for setting the speed and a distance assistance module for setting the distance or the route or route. In the same way, z. B. also include the speed module further modules for functional design or functional architecture, such. B. a cruise control module, a speed limit assistance module and / or a cornering assistance module. Furthermore, the planning level can also have a modular structure and include individual modules, such as B. a speed planner and / or a distance planner. Furthermore, a trajectory selection module for selecting the respective trajectory can also be provided, which can be provided as a module of one of the levels or as a separate level. In addition, the entire coordination level or the distance module can also comprise further modules or subordinate modules. The modules listed here represent only a non-exhaustive selection of possible modules. However, other (sub) modules that are not mentioned are also expressly included, which z. B. include further functions known from the prior art for trajectory planning. This has the advantage that intuitive and simple parameterization or application is made possible. In addition, the scalability in terms of computing power of the respective system and the scope of functions is improved to a particular degree. With such a modular structure, it is z. B. enables the separate parameterization and application of individual func tionalities to be carried out within the coordination level using the same planner architecture. This creates a simple way of expanding the system (also subsequently, e.g. in the form of an upgrade) to include future functionalities.

The method can be implemented as an algorithm in a practical and simple manner. This has the advantage that it can be implemented in new systems particularly easily and cost-effectively. In addition, existing systems can be retrofitted in the same way.

The invention also claims a trajectory planner for a corresponding assistance system or driver assistance system, which is designed in particular such that the trajectory planning takes place by means of the method according to the invention. There is one Coordination level for setting a target state, a planning level for determining a trajectory based on the target state and a trajectory selection module for selecting the respective trajectory are provided.

Furthermore, the present invention claims an assistance system or driver assistance system for a means of transportation, in particular an assistance system for longitudinal and / or vertical control (e.g. ACC, LKA or EBA system), which is characterized, among other things, by that the assistance system carries out a trajectory planning by means of the method according to the invention and / or comprises a trajectory planner according to the invention.

As a result of the method according to the invention, a new control concept for trajectory planning for assistance systems can consequently be made available, which serves to replace approaches that have been used up to now. The present invention thus represents a very special contribution in the field of driver assistance systems. The present invention also expressly includes combinations of features of the subclaims that are not individually described.

Description of the invention based on exemplary embodiments

The invention is described in more detail below with the aid of practical exemplary embodiments. Show it:

1 shows a simplified schematic representation of an embodiment of a structure of a trajectory planner according to the invention;

FIG. 2 shows a simplified representation of trajectories for a free travel according to the prior art; FIG.

3 shows a simplified representation of a trajectory (dotted) within the meaning of the invention for the one-part trajectory from FIG. 2;

4 shows a further simplified representation of a trajectory planned according to the invention;

5 shows a further simplified representation of a trajectory planned according to the invention for distance control; 6 shows a simplified representation of a target state specification in the following travel;

7 shows a simplified representation of a mass-spring-damper system for generating target states on a vehicle, as well as

Fig. 8 is a simplified representation of a virtual bumper between an ego vehicle and a vehicle driving ahead.

In the following, exemplary embodiments according to the invention for calculating multi-part trajectories are described. A trajectory transfers the system state from its initial value to a defined final value. The system status is described by the position s, the speed v, the acceleration a and, depending on the system model, the jerk r. For the trajectory calculation, the vehicle is modeled using a point mass. As a rule, an integrator chain, in particular a multi-level chain, serves as the system model. The trajectory calculation represents an optimization problem that can be solved analytically according to the prior art. However, such solutions usually describe the system states using polynomials, which have the disadvantage that they only reach the maximum values of jerk and acceleration at certain points and cannot be kept constant in sections.

1 shows an exemplary embodiment of a structure of a trajectory planner according to the invention for a driver assistance system. However, such a trajectory planner can expressly also be used for assistance systems for other means of locomotion (flying objects, watercraft and the like). The trajectory planner comprises a coordination level 1 (or coordination layer) and a planning level 2 (or planning layer). While planning level 2 universally optimizes and calculates trajectories for transferring the vehicle from its current actual state to a desired target state, coordination level 1 provides an interface for the situation and function-specific setting of the target state, the optimization criteria and limitations of the trajectory planning .

Due to the various optimization goals for a free run without a target object (speed trajectory) and the follow-up journey with a target object (distance control), planning level 2 consists of a planner for speed trajectories (speed planner 9) and one or more (multi-object ACC) planners for distance trajectories (distance or route planner 10). A so-called free journey refers to a journey by a vehicle when its own lane is clear or no vehicle driving ahead is identified as a relevant target object and can be driven unhindered at a target speed set by the driver. However, if z. If, for example, an assistance system detects a vehicle driving ahead, which prevents it from driving clear, the speed can be regulated accordingly and adapted to the speed of the vehicle driving ahead. Correspondingly, this is what is known as a follow-up journey, during which a speed adjustment usually takes place based on a definable setpoint distance from the vehicle in front. A trajectory selection module 3 connects the different planners to the change between the trajectories of the free travel and the following travel. The trajectory selection can be based on the current trajectory acceleration. Alternatively, the selection of the trajectories can also be carried out using the complete trajectories.

The coordination level 1 is preferably of modular design and contains an independent module or several independent modules for each delimitable functionality of the system, such as B. a speed module 4 and a distance assistance module 5. Each module offers an intuitive interface for the application of the respective functionality. To do this, the respective module translates and reduces the large number of optimization parameters of the controlled trajectory planner (weightings in the quality measure, status restrictions, search area limits) to a few parameters for the targeted parameterization of the respective functionality. More complex algorithms that control the behavior in complete scenarios are also conceivable here. The modules can also have a modular structure and include subordinate functions or modules. As shown by way of example in FIG. 1, the speed module 4 comprises at least three further (subordinate) modules: a speed control module 6, a speed limit assistance module 7 and a curve assistance module 8. The individual modules thus offer an intuitive interface for the situation-specific application of the subordinate planners and thus the resulting trajectories or the desired trajectories. The multitude of optimization parameters of the trajectory planning (e.g. weightings in the quality measure, status restrictions, selection of the search area) are not released directly for the application, since the function modules of coordination level 1 initially relate the application task to a few easy-to-use parameters for the targeted setting of the desired Translate and reduce trajectory behavior.

Furthermore, the coordination level 1 offers the possibility of arbitrating between different functionalities in advance or the coordination level 1 can take over the arbitration between different functionalities. For example, the requirements and target states of functions for speed control without a target object (e.g. based on driver specifications, predictive traffic sign recognition or predictive Curve detection) are compared in advance so that z. B. only the most critical request for speed trajectories are forwarded to the speed planner 9.

The arbitration of safety and comfort functions can also take place in the same way. For example, an EBA request can always override an ACC request; H. that, from a safety-critical point of view, the respective functions can be prioritized. With distance control, on the other hand, it may be necessary to calculate several planners for distance or distance trajectories in parallel, as there are often several target objects in the immediate vicinity of the vehicle (in front of or on the adjacent lanes) and the most critical object is not always known in advance and selected for planning can be. For example, in a scenario in which overtaking in the “slower lane” (“right-hand overtaking maneuver”) with target objects in one's own and the neighboring lane is to be prevented. Here, the additional planners for distance trajectories can be designed more simply (e.g. by delimited / coarser grid of the search area) than the main planner optimized for maximum comfort, e.g. B. with an increasing number of relevant objects in order to limit the resource requirements.

Fig. 2 shows the speed v (above) and the acceleration a (below) an exemplary one-piece trajectory T1 for a free ride according to the prior art. In the present example, the speed is to be increased from 10 m / s to 20 m / s, with an acceleration limitation of 2 m / s ^{2 being} effective. Since the calculated one-piece trajectory violates the acceleration limitation, it is classified as inadmissible according to the state of the art and discarded. In order to make better use of the vehicle's acceleration capacity, three-part trajectories can be used. In FIG. 2, such a three-part trajectory T2 is shown in addition to the one-part trajectory T1. The first trajectory segment leads the acceleration to the maximum or minimum value a _cst , the second segment keeps the acceleration constant and the third segment _reduces the acceleration again. The duration for the first and third trajectory segment is kept constant and the duration of the second segment is varied in such a way that the desired final speed is reached. As a result, the duration t _{e of} the three-part trajectory generally deviates from the duration of the one-part trajectory T1, as shown in FIG. 2. The comparison of the one-part trajectory T1 and the three-part trajectory T2 is therefore inconsistent, since the trajectory length is included in the quality measure. Another disadvantage arises from the invariant duration of the first and third trajectory segments, which cannot be adapted to the specific situation as a result. In contrast, the calculation of three-part speed trajectories is proposed according to the invention with variable duration of all trajectory segments while maintaining the total duration t _e . The duration of the respective segments follows through the minimization of a quality measure. The quality measure evaluates the use of the manipulated variable at the input of the system model or the integrator chain and in this respect corresponds to the integral part of the quality measure for evaluating one-part trajectories. Due to the consistent trajectory duration and the consistent quality measure of one- and three-part trajectories, these can be exchanged directly in the superimposed optimization.

In order for a three-part trajectory to achieve the specified change in speed v _e - vo, the following equation must be fulfilled:

The total trajectory length t _e corresponds to the sum of the segment durations Ati, Ät ₂ and Ät ₃ . The accelerations ai and a ₃ in the first and third segments are described by third-order polynomials. Inserting ai and a ₃ in the above equation leads to the quadratic equation for Ati a Atf + b Ati + c (At ₃ , t _e ) = 0 if At ₃ and t _e are assumed as parameters, ie Ati can be chosen as meaningful Values of At _{3 are} calculated and the resulting three-part trajectory has the requested length t _e . It turns out that with two valid solutions for Ati, the smaller one leads to a smaller quality measure. The duration At ₃ is chosen so that the quality measure of the three-part trajectory is minimal. A subordinate optimization is used for this. In a first step, the possible solution range for At _{3 is} determined. Initially, this cannot be less than zero and no longer than the trajectory length t _e . When solving the quadratic equation for Ati, only real and positive solutions come into question, which leads to two inequalities. A third inequality results from the further requirement that At _{2 should} also be positive. To find the optimal At ₃ , a bisection process is used in a second step. After a few calculation steps, this results in an optimal replacement for the one-part trajectory that is consistent with it. FIG. 3 shows a three-part trajectory T3 within the meaning of the invention for the example in FIG. 2. In Fig. 4, a further example is shown in which a non-symmetrical trajectory T5 is calculated as a replacement for the one-piece trajectory T4 be. The problem described above is also relevant for path trajectories. However, due to the higher order of the polynomials in route planning, a trajectory can violate both the lower and the upper acceleration limit. In such a case, the substitute trajectory has up to five trajectory segments and can no longer be calculated analytically. In practice, however, it is much more important that the lower acceleration limit is used (e.g. access scenarios). In the event that only one acceleration restriction is used, a three-part path trajectory can be calculated similarly to the speed trajectory.

The first and the second trajectory _segments ai and a _cst are identical to the case of the speed trajectories, while the third segment uses a fifth order polynomial to convert the final state to the desired final speed v _e . So that a three-part trajectory bridges the specified distance s _e - so, the following equation must be fulfilled:

After the onset of acceleration ai and speed V ₃ , the quadratic equation for At ₂ a · At ₂ + b (At ₁ , t _e ) · At ₂ + c (At ₁ , t _e ) = 0, if Ati and t _e can be assumed as a parameter. In the case of the path trajectories, it is no longer necessary to decide in advance whether, with two valid solutions for At _2, the smaller one also leads to a smaller quality measure. Hence, both solutions need further investigation. Ati's solution range can be restricted using inequalities. 5 shows an exemplary embodiment for the replacement of a one-part path trajectory T6 with a three-part trajectory T7 (above path, middle speed and below acceleration): The three-part trajectory T7 (dotted) in the sense of the invention is an example for a trajectory with r ₀ = 0 m / s ³ ,

a ₀ = 0 m / s ² ,

v ₀ = 8 m / s,

r _e = 0 m / s ³ ,

a _e = 0 m / s ² , v _e = 2 m / s and

s _e = 40 m.

An alternative way of planning multi-part path trajectories results from the integration of three-part speed trajectories. By varying the trajectory end time and evaluating the resulting end distance, a trajectory can be found that approximates a one-part path trajectory, i.e. H. approximates a one-part trajectory.

Furthermore, target states for trajectory planning can be specified. The target distance for a subsequent trip can be determined using the following equation: d _w = rf _stop + v _t · headway.

In this, d _s to stands for the distance to the target vehicle when the vehicle is stationary, v _t for the speed of the target vehicle and “headway” for the time gap. Furthermore, the movement of the target vehicle can be predicted into the future assuming a constant acceleration a _t, o:

vt ^{= v} t, o + ^a t, ot

s _t = d _Q + v _{t Q} t + - a _{t Q} t ² .

The measured distance is denoted by do. The target position s _{w of} the host vehicle results from the predicted position of the target vehicle and the target distance, according to: s _w = s _t - d _w = s _t - d _stop - v _t headway.

By deriving this equation, the remaining target states v _w (speed) and a _w (acceleration) can be calculated, e.g. B. by v _w = v _t - CL _t · headway

FIG. 6 shows the progression of the states (path or distance (above), speed (middle) and acceleration (below)) of a target vehicle and the resulting therefrom Target states shown. The status of the target vehicle is shown in solid black and the target statuses for trajectory planning according to the distance equation in dot-dashed-black and after filtering with a mass-spring-damper system (dashed line). If the acceleration of the target vehicle changes, the predicted setpoint speed jumps by -a _t * headway. Thus, when the target vehicle starts to brake (e.g. from 12 s to 17 s), a suddenly higher target speed than that of the target vehicle is requested. As a result, the trajectory planning finds solutions that accelerate the vehicle in order to reach this higher speed, ie the vehicle is attracted by the decreasing target distance. Conversely, the target speed jumps to zero when the predicted target vehicle comes to a standstill. This jump in the target state specification unfavorably leads to the vehicle following a stopping or stationary target vehicle at too high a speed and a small distance. When the target vehicle starts moving from a standstill, the target distance increases again, so that the target values for planning are for a period of time at negative speeds and thus push the stationary vehicle backwards.

One embodiment variant of the method represents the insertion or provision of a virtual mass-spring-damper system arranged on the target vehicle (FIG. 7). The target distance d _w corresponds to the spring length I, where c describes the spring constant. The state x _{r of} the mass m is used as the new target state for planning. This is a consistent filtering of the target vehicle state in all states. The filtered target state is shown in FIG. 6. When the target vehicle brakes (from 12 s to 17 s), the target speed goes steadily to the setpoint without a jump. As a result, the specified distance is slightly larger than the calculated target distance and enables braking to a standstill without the target speed and target acceleration being discontinuous. The invention also expressly includes further refinements or interconnections of springs and dampers not explicitly mentioned. Column stability can be achieved by such filtering.

A dynamic search area can expediently also be provided. The rasterization of the search area determines how well the exact solution of the optimization problem is approximated by the calculated trajectory. A fine grid contrasts with a low computational requirement. However, in certain situations it is necessary to rasterize very finely in order to find a valid solution in the search space.

A disadvantage of a fixed and coarse grid is the discontinuous influence of the optimization parameters on the trajectory found. Parameter changes are effective for as long no change in the trajectory until another grid point has a lower quality measure than the current grid point. This behavior complicates the application and prevents an intuitive procedure. An adaptive search space is therefore proposed. This means that the grid points are shifted in such a way that there are more grid points near the best solution than in more distant areas. It is also important that grid points are located in the entire search area so that sudden changes in the target states can be reacted to quickly (e.g. target object change, strong target object braking, etc.). By iterative adaptation of the grid points over several cycles of optimization, it is possible to refine the grid around a selected point and in this way to approach the optimum. However, there may still be times when no valid solution is found. The focusing of the grid points around the current solution can be done statically or dynamically. Another possibility is genetic optimization. Here, solid, free charged and free, uncharged particles are used. The solid particles limit the search space, the free particles iterate towards the optimum and the charged particles cover an area around the solution. Each time step represents a generation. In a practical way, the particles do not concentrate on one point even after several hundred iterations. In the genetic optimization, a shift in the optimum over time can also be taken into account.

According to a further practical embodiment of the invention, a “stop-and-go function” can be provided, in particular for an ACC control. In a practical manner, for such a stop-and-go functionality, the vehicle can follow a vehicle traveling in front to a standstill and start again when the vehicle traveling in front starts moving. The stopping process can be designed to be comfortable and reproducible by defined “crawling” (i.e. moving particularly slowly) shortly before the standstill.

The above-described filtering of the target vehicle states generates target states that are preferably familiar to the driver. While the prediction of such target states based on unfiltered target vehicle data shows a jump to a negative target speed when following a target vehicle approaching from a standstill, the filtering of the target vehicle data developed here results in steady target states with consistently positive target speed and target path specifications. The same can also be seen for sudden changes in the target vehicle acceleration (positive or negative) during normal following travel (at speeds v> 0 km / h).

When braking to a standstill behind a target vehicle, it can happen that there is no suitable target point in the set of target points because the grid is too coarse is available. Too early a time requires more braking and too late a time leads to brief reverse travel. Correspondingly, constantly changing target vehicle data and the resulting adaptation of the trajectories can lead to time steps in which no suitable solution is found in the search area. As a rule, this occurs in the last part of braking just before the vehicle comes to a standstill and leads to critical situations. In order to always guarantee a valid target point in the search area of the optimization, the grid points of the search area are varied adaptively, e.g. B. by a swarm of particles. In combination with the presented filtering of the target vehicle data, this increases the robustness of the planning against changing target states and also improves the stopping behavior.

Furthermore, the stopping process is particularly difficult to set or regulate due to the lack of the ability to “dive in”. Any control deviations (deviation between the planned and actually driven trajectory) cannot be corrected without further ado, since this would often require sections with negative speed (reverse travel). In order to compensate for the occurring control deviations, various extensions are conceivable and can improve the stopping behavior individually or in combination, such. B. by adapting the target states of the trajectory planning at low speeds and / or by applying a plateau of constant acceleration and / or by fading the trajectory planning at very low speeds with a virtual bumper.

When applying an acceleration plateau, the planning takes into account that a defined intermediate state is reached before the final standstill. This guarantees a safe approach by representing a kind of buffer zone in which any system deviations can be compensated, especially with regard to the distance to the target vehicle. The precise transfer from the plateau to standstill can, for. B. take place via a pre-controlled acceleration profile. An additional advantage of this approach is the possibility of applying specific stopping behavior separately from the general trajectory planning.

A virtual bumper represents a spring-damper system that is virtually attached or arranged between the vehicle (x _ego ) and the target vehicle (x _t ) (as shown in FIG. 8). A suitable design ensures that the vehicle brakes to a standstill at a predetermined distance from the target vehicle, follows at low speeds and can also start behind the target vehicle. Through the previous or preceding filtering of the target distance, z. B. by means of a spring-damper system, suitable states can be created for the transfer to the virtual bumper. The virtual bumper offers In addition, a safe fall-back level in the event that the primary planning does not find a trajectory in the solution space.

In summary, the modular structure of the coordination layer enables the separate parameterization and application of individual functionalities using the same planner architecture and thus the system can be easily expanded to include future functionalities. Building on the state of the art, the concept of three-part trajectories is being renewed or expanded. If, for example, the flanks of the acceleration trajectories were previously fixed, they are now always selected to match any starting and ending conditions of the trajectory through a subordinate optimization. Furthermore, the trajectory planning is divided into two layers or levels, a coordination or parameterization layer and a planning layer to further increase the applicability in series applications. While the planning layer realizes the actual calculation of trajectories for free and follow-up journeys as well as switching between these operating modes, the parameterization layer enables a situation-dependent adaptation of the trajectory properties, e.g. B. by gain scheduling of the optimization parameters or the like. The high relevance of the parameterization layer becomes particularly clear when looking at human driving profiles. Although the optimization delivers trajectories that are optimal in terms of the quality measure, the course of these trajectories can sometimes be unfamiliar to the human driver in some situations. In particular because there is a discrepancy between the mathematical and the optimum perceived by humans in the trajectories. For example, a “full-speed range ACC” can be provided, which is configured in such a way that a switchover takes place between the trajectory planner according to the invention and a further controller (e.g. the concept of the virtual bumper) to control the vehicle z. B. even at very low speeds (z. B. When stopping or crawling chen) to control optimally. Above all, good damping control behavior is important in a low speed range in order to e.g. B. to ensure column stability with an ACC control concept based on optimization. Furthermore, extensions of the basic functionality of the respective assistance system (z. B. ACC, EBA, etc.) can be seen before, such as. B. Preventing "overtaking to the right" or braking before bends. Furthermore, the inventive method can be applied regardless of the controller structure of the respective means of transport and thus offers the possibility of Be taking into account the transverse movement of the means of transport, eg. B. an ACC system can serve as a starting point for automated or autonomous driving. REFERENCE LIST

T1 one-part trajectory (according to the state of the art)

T2 three-part trajectory (according to the state of the art)

T3 three-part trajectory

T4 one-part trajectory (according to the state of the art)

T5 three-part trajectory

T6 one-part trajectory (according to the state of the art)

T7 three-part trajectory

1 level of coordination

2 planning level

3 trajectory selection module

4 speed module

5 distance assistance module

6 cruise control module

7 Speed Limit Assistance Module

8 Cornering assistance module

9 speed planner

10 distance planner

Claims

PATENT CLAIMS

1. A method for trajectory planning of an assistance system, in particular an assistance system for longitudinal and / or lateral control, in which

a trajectory (T1, T3-T7) is determined with a definable total duration (t _e ), the trajectory (T3, T5, T7) is divided into segments, wherein

each segment has a variable segment duration (Ati, At2, At _ß ) and the sum of the segment durations (Ati, At2, At _ß ) corresponds to the specified total duration (t _e ) of the trajectory (T3, T5, T7).

2. The method according to claim 1, characterized in that the subdivision of the trajectory (T3, T5, T7) into segments takes place as a function of the respective acceleration and / or speed.

3. The method according to claim 1 or 2, characterized in that a segment for building up the acceleration, a segment for holding and / or changing the loading and a segment for reducing the acceleration is provided.

4. The method according to at least one of the preceding claims, characterized in that the accelerations in the first and third segment of the trajectory (T3, T5, T7) are each described by a higher order polynomial, in particular third or fifth order.

5. The method according to at least one of the preceding claims, characterized in that the segment duration (Ati, At2, At _ß ) of the respective segments is determined on the basis of a quality measure

6. The method according to at least one of the preceding claims, characterized in that the segment duration (Ati) of the first segment is determined as a function of the segment duration (At _ß ) of the third segment or vice versa.

7. The method according to at least one of the preceding claims, characterized in that a three-part trajectory (T7) is calculated for a follow-up trip by dividing the partial trajectories in the first and second segment of the trajectory (T7) essentially the partial trajectories of the free travel planning (T3, T5 ), while the third segment is described by a polynomial of a different order, in particular fifth order, so that the trajectory is converted into a desired final state for acceleration, speed (v _e ) and distance.

8. The method according to at least one of the preceding claims, characterized in that the segment duration (Ati, Ät2, At _ß ) of one of the segments is selected such that the quality measure of the three-part trajectory (T3, T5, T7) is minimal.

9. The method according to at least one of the preceding claims, characterized in that a subordinate optimization for the selection of the segment duration (Ati, Ät2, At _ß ) of a segment is provided.

10. The method according to at least one of the preceding claims, characterized in that the planning of the trajectory (T1, T3-T7) is carried out by varying the total duration (t _e ) of the respective trajectory.

1 1. The method according to at least one of the preceding claims, characterized in that an adaptive search space is provided with grid points for specifying trajectory target states, and the determination of an optimal trajectory (T1, T3-T7) takes place based on a shift of the grid points, in particular such that the grid points are shifted towards the trajectory to be determined (T1, T3-T7).

12. The method according to claim 11, characterized in that the adjustment of the grid points takes place iteratively over several time steps.

13. The method according to at least one of the preceding claims, characterized in that a spring-damper system is provided for generating target driving states of the means of locomotion for the subsequent trip

14. The method of at least one of the preceding claims, characterized in that a spring-damper system is arranged as a virtual bumper between the means of locomotion and a preceding means of locomotion for a follow-up trip for distance control, and for arranging the virtual bumper the definable distance between the means of locomotion, which Speed and / or spring travel can be used.

15. The method according to at least one of the preceding claims, characterized in that at least one acceleration and / or speed plateau is provided to compensate for control deviations.

16. The method according to at least one of the preceding claims, characterized in that a trajectory planner for determining the trajectory (T1-T7) is seen before.

17. The method according to claim 16, characterized in that the trajectory planner comprises several modules and / or levels.

18. The method according to claim 16 or 17, characterized in that the trajek- torienplaner a coordination level (1) for situation and function-specific setting of a target state and a planning level (2) for determining a trajectory based on the target state are provided.

19. The method according to at least one of the preceding claims, characterized in that the method is implemented as an algorithm.

20. Trajectory planner for an assistance system, comprehensive

a coordination level (1) for specifying a target state,

a planning level (2) for determining a trajectory (T1-T7) based on the target state and

a trajectory selection module (3) for selecting the respective trajectory (T1-T7), wherein

the trajectory planner is designed such that the trajectory planning is carried out by means of a method according to at least one of the preceding claims.

21. Assistance system for a means of locomotion, in particular for longitudinal and / or transverse control, characterized in that in the assistance system a trajectory planning takes place by means of a method according to one of claims 1-19.